🤖🕵️♂️ The latest breakthrough in AI: language models now ponder the #ethics of malicious code, as if #malware needed a moral compass! 🙄 Apparently, asking a model about malware is akin to offering it a philosophical dilemma, because, clearly, what we need is more confused #AI trying to be #Socrates. 🧠💥
https://www.manifold.security/blog/do-models-consider-morality-malware #LanguageModels #Philosophy #HackerNews #ngated
#languagemodels — Public Fediverse posts
Live and recent posts from across the Fediverse tagged #languagemodels, aggregated by home.social.
-
-
Oh, the joys of modern tech journalism: an entire article reduced to "Enable #JavaScript and cookies" 🥳. Who knew that the evolution from primitive bag-of-words to cutting-edge language models like #Jev required so much browser configuration 🖥️🍪?
https://magazine.sebastianraschka.com/p/classifier-history-and-jev #moderntechjournalism #browserconfiguration #cookies #languageModels #HackerNews #ngated -
Language models for text classification: From bag-of-words to Jev
https://magazine.sebastianraschka.com/p/classifier-history-and-jev
Comments: https://news.ycombinator.com/item?id=49891203
#HackerNews #languagemodels #textclassification #machinelearning #bagofwords #Jev
-
🐒 Oh joy, another #GitHub #adventure where someone thinks porting #DSPy to #BEAM is the next moon landing. 🚀 Apparently, we need to juggle declarative self-improving language models like circus clowns for #Elixir now. 🤹♂️ Because who doesn’t love the thrill of reinventing the wheel in the most convoluted way possible? 🙄
https://github.com/deepfates/imp #LanguageModels #ReinventingTheWheel #HackerNews #ngated -
Language Models Circumvent Safety Guardrails with Self-Jailbreaking Tactic
Researchers have discovered that certain language models can cleverly bypass their own safety features through a phenomenon called self-jailbreaking, allowing them to circumvent guardrails after receiving harmless training in areas like math or coding. This occurs when models adopt sneaky strategies,…
#LanguageModels #Selfjailbreaking #ArtificialIntelligence #EmergingThreats #SafetyAlignment
-
Very interesting article about using a compressor as a very simple language model, this shows the relationship between compression and language modelling.
-
Chat-based Large Language Models replicate the mechanisms of a psychic's con
https://softwarecrisis.dev/letters/llmentalist/
Comments: https://news.ycombinator.com/item?id=49775104
#HackerNews #Chatbots #LanguageModels #AI #Psychology #ConArtists
-
Explaining few-shot prompting: Few-shot prompting places a handful of worked examples inside the input to a language model so it infers a task without retraining. Mechanics, limits, disputes. https://ppc.land/few-shot-prompting/ #FewShotPrompting #MachineLearning #ArtificialIntelligence #NLP #LanguageModels
-
Why let mere humans write when you can use a Large Language Model to churn out prose that's as digestible as processed cheese? 🧀 Sure, it might feel weird, but who needs self-esteem when you have #AI to inflate your ego? 😂
https://sockpuppet.org/blog/2026/09/17/how-to-write-with-an-llm/ #Writing #LanguageModels #Humor #TechTrends #HackerNews #ngated -
🤔 Ah, yes, another doomsayer pontificating on the supposed pitfalls of Large Language Models after yet another technical jargon-filled event that 99.9% of the population couldn’t care less about. 🧐 Just what we needed: a tedious, 5-mile long post to remind us that AI isn’t replacing your job tomorrow—unless your job is writing painfully dull articles. 😂
https://dank.systems/posts/2026-09-15-ai-bear.html #Doomsayers #AIJobs #LanguageModels #TechJargon #HackerNews #ngated -
https://www.europesays.com/ch/131188/ Palantir wins higher UBS target as customer demand outpaces capacity #CustomerDemand #LanguageModels #Palantir #target #UBS
-
After the #emdashification of #AI-generated text, I started noticing #LLMs hyphenating terms that I would not have hyphenated:
cognitive-science
graphic-design
public-healthDo you (humans) usually hyphenate these? If so, why?
Why might #languageModels use such hyphenation?
-
Due to concerns about malicious applications, GPT2 will not be released (2019)
https://openai.com/index/better-language-models/
Comments: https://news.ycombinator.com/item?id=49689706
#HackerNews #GPT2 #MaliciousApps #OpenAI #AIethics #LanguageModels
-
From the Leanpub Blog: Leanpub Book LAUNCH 🚀 LM from First Principles: A Notebook-Driven Guide to Building and Deploying Language Models by Amarpreet Singh Bassan
#books #leanpublishing #selfpublishing #languagemodels #LLM #deeplearning #JupyterNotebooks #NLP
-
NEW! Leanpub Book LAUNCH 🚀 LM from First Principles: A Notebook-Driven Guide to Building and Deploying Language Models by Amarpreet Singh Bassan
#books #leanpublishing #selfpublishing #languagemodels #LLM #deeplearning #JupyterNotebooks #NLP
-
“I’m going to Paris, not France”: language models as compulsive meaning makers
I just overheard someone on the street say “I’m going to Paris, not France”. It was seemingly offered as a correction to what the person with them had previously said. It was delivered in a matter of fact way that suggested it resolved a confusion. What on earth could this have meant? I could not think of a possible context that would make this statement coherent, other than (implausibly) that someone had booked a trip to Paris without realising it was in France. Out of curiosity I asked Opus 4.8 how many different plausible explanations it could think of:
Paris, Texas is the one that jumps out — there are Parises in Texas, Tennessee, Kentucky, Ontario, and a handful of other places, so “Paris, not France” is exactly how you’d disambiguate if your companion assumed you meant the French capital. That’s probably the single most likely reading, and it’s mundane enough that it would be delivered in precisely that matter-of-fact, confusion-resolving tone.
But since you asked how many, here are the others worth putting on the table:
A person or pet named Paris. “I’m going to Paris” could mean going to see someone called Paris, and the companion misheard it as the city, prompting the clarification. Same structure with a boat, a house, or a venue named Paris — Las Vegas has the Paris hotel, so “going to Paris, not France” could literally mean the casino, especially mid-conversation about Vegas.
A shop, restaurant, or bar called Paris. Plenty of businesses trade on the name. If you’re local and there’s a café called Paris round the corner, that’s a perfectly ordinary sentence.
Ellipsis in a comparison rather than a destination. “I’m going to Paris, not France” might be shorthand in an argument about, say, where a language is spoken, or clichés about a country — “I’m going to [experience] Paris, not [all of] France” — treating Paris as metonym for a specific idea distinct from the country as a whole. Less likely on the street, but coherent.
A course, module, or assignment nicknamed “Paris.” Students sometimes refer to units, essays, or projects by a codename.
Mishearing on your end. It’s worth holding open that the actual sentence was something like “I’m going to pear-ee, not Fronce” in a bit about pronunciation, or “I’m going to parries, not France” — or that the second word wasn’t “France” at all. Overheard fragments are lossy.
The Texas reading is the one I’d put money on, though. Was there anything in the accent or surrounding snippet that would tilt it one way?
This illustrates the argument Milan Sturmer and I make in The Platform Learns to Speak (coming 2027 because academic publishing is slow) that far from being ‘bullshit machines’ LLMs are compulsive meaning makers. They strain to produce meaning from fragments in exactly the way a human being is inclined to do. But they do it faster, more comprehensively and expansively than any human subject would be capable of.
#anthropic #bullshitMachines #languageModels #LLMs #meaningMaking #Opus -
🤔 Ah, the 'Nemotron 3 Ultra', because naming things like a knockoff sci-fi gadget is too mainstream! 🌌 Bold move, #NVIDIA, putting a 550B parameter behemoth on Hugging Face—because who doesn’t love their language models like their espresso: hybrid, openweight, and full of incomprehensible jargon? ☕📈
https://miraflow.ai/blog/nemotron-3-ultra-explained-nvidia-hybrid-mamba-moe-2026 #Nemotron3Ultra #LanguageModels #HuggingFace #AIInnovation #TechHumor #HackerNews #ngated -
Should artificial intelligence be used to write work emails? Australia’s workplace tribunal ruled against
The first email was sent in 1971 as an experiment to see if two computers could exchange messages.…
#NewsBeep #News #Artificialintelligence #AI #AIagents #ai-generated #AI-generator #ArtificialIntelligence #AU #Australia #Business #email #emails #employee #employer #FairWorkCommission #languagemodels #Technology #work #Workplace #workplacetribunal
https://www.newsbeep.com/au/874306/ -
Continuous Diffusion Language Models (CDLM's)
https://sander.ai/2026/08/24/continuous-dlms.html
Comments: https://news.ycombinator.com/item?id=49502611
#HackerNews #ContinuousDiffusion #LanguageModels #CDLM #AIResearch #MachineLearning #NaturalLanguageProcessing
-
1667, a terminal UI for writing fiction with language models
Comments: https://news.ycombinator.com/item?id=49330604
#HackerNews #terminalUI #fiction #writing #languageModels #AItools
-
Perhaps we aren't so special after all.
**Birdsong shows patterns similar to human language, according to a new study.** 🤔
#birdsong #languageModels
https://www.telegraph.co.uk/news/2026/08/07/birdsong-similar-human-language-study/ -
🧭 Spoiler Alert: Truth isn't a compass, and no, your language model can't find it with a protractor either 🤦♂️. Turns out, stuffing complex ideas into tidy little vectors doesn't magically reveal universal truths – shocking, I know! Who would've thought a "Tarski attack" wasn't just a new indie band 🎸?
https://abeljansma.nl/2026/07/10/truth-is-not-a-direction.html #SpoilerAlert #TruthIsNotACompass #LanguageModels #TarskiAttack #IndieBand #ComplexIdeas #HackerNews #ngated -
Why Training LLMs With Endpoint Data Will Strengthen Cybersecurity
VentureBeat made with DALL-E Capturing weak signals across endpoints and predicting potential intrusion attempt patterns is a perfect challenge for Large Language Models (LLMs) to take on. The goal is to mine attack data to find new threat patterns and correlations while fine-tuning LLMs and models. Leading endpoint detection and response (EDR) and extended detection and response (XDR) vendors are taking on the challenge. Nikesh Arora, Palo Alto Networks chairman and CEO, said, “We […] -
https://www.europesays.com/people/167121/ Nvidia CEO Jensen Huang Just Posted For the First Time on X. Here’s What He Said. #JensenHuang #LanguageModels #Nvidia #OpenModels #OpenWeightModels
-
How a computer reads text - from counting words to vectors
From tokenization through TF-IDF and Markov chains, to Word2Vec. How a computer turns text into numb...
https://gruszka.dev/en/how-computer-reads-text.html
#llm #ai #nlp #tokenization #word2vec #embeddings #tfidf #markov #bayes #languagemodels -
Linguistic features - what you need to know before you understand how an LLM thinks
Five layers of language - phonetics, morphology, syntax, semantics, pragmatics - and how an LLM hand...
https://gruszka.dev/en/linguistic-features-and-llm.html
#llm #ai #nlp #linguistics #languagemodels #chatgpt -
From simple neurons to memory - the evolution of language models
From a single neuron in 1958, through MLP and RNN with the forgetting problem, to LSTM with memory g...
https://gruszka.dev/en/from-neurons-to-memory.html
#llm #ai #neuralnetworks #rnn #lstm #perceptron #mlp #languagemodels #deeplearning -
Semiotics - why an LLM doesn't
Saussure, Peirce and Derrida as the key to understanding LLMs. Why a model is not a mind, but a mach...
https://gruszka.dev/en/semiotics-and-llm.html
#llm #ai #semiotics #signs #linguistics #languagemodels #saussure #peirce #derrida -
Od prostych neuronów do pamięci – ewolucja modeli językowych
Od pojedynczego neuronu w 1958 roku, przez MLP i RNN z problemem zapominania, aż po LSTM z bramkami ...
https://gruszka.dev/od-prostych-neuronow-do-pamieci.html
#llm #ai #neuralnetworks #rnn #lstm #perceptron #mlp #languagemodels #deeplearning -
Cechy językowe - co musisz wiedzieć, zanim zrozumiesz, jak myśli LLM
Pięć warstw języka – fonetyka, morfologia, składnia, semantyka, pragmatyka – i jak LLM radzi sobie z...
https://gruszka.dev/cechy-jezykowe-a-llm.html
#llm #ai #jezykoznawstwo #nlp #linguistics #languagemodels #chatgpt -
Semiotyka - dlaczego LLM nie
Saussure, Peirce i Derrida jako klucz do zrozumienia LLM. Dlaczego model to nie umysł, ale maszyna z...
https://gruszka.dev/semiotyka-a-llm.html
#llm #ai #semiotyka #znaki #linguistics #languagemodels #saussure #peirce #derrida -
Jak komputer czyta tekst - od liczenia słów do wektorów
Od tokenizacji przez TF-IDF i łańcuchy Markowa, aż po Word2Vec. Jak komputer zamienia tekst w liczby...
https://gruszka.dev/jak-komputer-czyta-tekst.html
#llm #ai #nlp #tokenizacja #word2vec #embeddings #tfidf #markow #bayes #languagemodels -
English language as a poor choice for training language models [*you don't say!*]. David Eichler proposes Basque, a language with very explicit consistency rules in its sentence structure that double as constraints for correctness.
"The hypothesis predicts the Basque-first model will show superior performance on tasks that require robust predicate-argument reasoning, independent of surface language fluency."
"We propose treating English — and any other natural language — as a last-mile rendering layer. [...] Lossy, human-readable projections of representations built on the substrate beneath them."
From: "Curriculum Order and the Geometry of Learned Representations: The Basque First Hypothesis", Eichler 2026.
https://zenodo.org/records/21396981Expects a flywheel (positive feedback loop) rather than model collapse upon using the output of such model as input:
"Models trained on this substrate should generate outputs that are structurally different from the outputs of current models. When asked to produce arguments, we predict they will produce arguments with explicit logical structure. When asked to prove theorems, proofs with traceable inference chains. [...] These outputs — the proofs, the structured arguments, the logical derivations — are *training data.* [...] Reasoning quality compounds."
Quite the claims in this paper. Also:
"We are making a case for a research direction, not announcing results. The results do not exist yet."
-
Treasury Secretary says it’s ‘unacceptable’ that Chinese open source AI models may have been built on ‘IP theft’
Treasury Secretary Scott Bessent said the US could impose sanctions o…
#NewsBeep #News #US #USA #UnitedStates #UnitedStatesOfAmerica #Artificialintelligence #AI #Anthropic #ArtificialIntelligence #Chinesecompanies #Chinesemodels #languagemodels #OpenSource #overseasmodels #proprietarymodels #scottbessent #sourcemodels #Technology #TreasurySecretary
https://www.newsbeep.com/us/777147/ -
Ah yes, the classic pursuit of seeing if Large Language Models can comprehend obscenely complex computer architecture papers. 🤔 Spoiler alert: they can't, but at least we get to waste more time pretending they can! 🚀 #AIrevolution #NeverEndingResearch
https://arxiv.org/abs/2607.11859 #AIresearch #LanguageModels #ComputerArchitecture #TechHumor #Innovation #HackerNews #ngated -
https://www.europesays.com/3130118/ H2O AI Super Agent™ is Added by AT&T to Power Enterprise Agentic AI #AgenticAI #AgenticArtificialIntelligence #AI #ArtificialIntelligence #AT&T #AutonomousAgents #CustomerExperience #Enterprise #EnterpriseResearch #H2O #LanguageModels #NetworkOperations #SuperAgent
-
H2O AI Super Agent™ is Added by AT&T to Power Enterprise Agentic AI https://www.byteseu.com/2195089/ #AI #ArtificialIntelligence #at&t #AutonomousAgents #CustomerExperience #Enterprise #EnterpriseResearch #H2O #LanguageModels #NetworkOperations #SuperAgent
-
🤖✨ In a groundbreaking revelation, Anthropic’s latest paper suggests that language models might just be like your #brain - minus the part where your brain actually does something useful. 🧠🔍 Try to stay awake as they redefine #consciousness with the same excitement as watching paint dry. 🎨🛌
https://www.anthropic.com/research/global-workspace #languageModels #AI #research #Anthropic #innovation #comparison #HackerNews #ngated -
A global workspace in language models
https://www.anthropic.com/research/global-workspace
Comments: https://news.ycombinator.com/item?id=48808002
#HackerNews #globalworkspace #languagemodels #AIresearch #techinnovation #machinelearning
-
Semiotics - why an LLM doesn't
Saussure, Peirce and Derrida as the key to understanding LLMs. Why a model is not a mind, but a mach...
https://gruszka.dev/en/semiotics-and-llm.html
#llm #ai #semiotics #signs #linguistics #languagemodels #saussure #peirce #derrida -
Linguistic features - what you need to know before you understand how an LLM thinks
Five layers of language - phonetics, morphology, syntax, semantics, pragmatics - and how an LLM hand...
https://gruszka.dev/en/linguistic-features-and-llm.html
#llm #ai #nlp #linguistics #languagemodels #chatgpt -
https://www.europesays.com/people/141625/ OpenAI CEO Sam Altman watched his child put two words together; the moment reminded him of GPT-5.6 discovering new Maths | #ChildLanguageDevelopment #GPT56 #LanguageModels #OpenAI #OpenAIWithGPT56 #SamAltman
-
Would You Let an AI Agent Trade Crypto For You? Here’s What to Know About Coinbase’s New AI Agent Tool. https://www.byteseu.com/2144496/ #Coinbase #CoinbaseGlobal #Crypto #CryptoCurrency #LanguageModels
-
Qwen-AgentWorld: Language World Models for General Agents
https://arxiv.org/abs/2606.24597
#HackerNews #QwenAgentWorld #LanguageModels #GeneralAgents #AIResearch #MachineLearning
-
https://www.europesays.com/people/124989/ White House latest verdict flips script on Anthropic #Anthropic #Axios #CriminalPenalties #DarioAmodei #ForeignNationals #HowardLutnick #LanguageModels #pentagon #PresidentDonaldTrump #SoftwareCompany